1. Test indicators

1.1 Average response time ART：

Response time follows 2、5、8s principle , The response time of this test is less than or equal to 8s;

1.2 Number of concurrent users ：

Now the number of operators on peak days 500 people ,20% Calculation of concurrency , The number of concurrent users on peak days is greater than or equal to 100

Planning for the future 2 The annual peak operation reaches 600 people ,20% The concurrency value , The number of concurrent users is greater than or equal to 120

It is planned that the number of operators will reach... In the next three years 700 people ,20% The concurrency value , The number of users is greater than or equal to 140.

1.3 Resource usage indicators ：

cpu The utilization rate is less than or equal to 80%

Memory utilization is less than or equal to 80%

The disk exchange rate is less than or equal to 80%

1.4 tps value ：

Number of transactions processed per second ,80% My deal is in 20% Time to finish , Daily trading volume 10 Ten thousand brush , One day 8 Hours

tps=80%*100000/（8*3600*20%）=13.89

Concurrent transaction success rate ： Greater than or equal to 95%

2. Test type

2.1 The benchmark

Single transaction single user test , It takes time for a typical transaction to obtain a single transaction without pressure , Provide a data reference for subsequent concurrent testing , A user runs 5 minute .

Verify the correctness of test script and test parameters .

Get the performance data of a single transaction , It is mainly the average response time of a single transaction 、TPS.

2.2 Concurrent test

It is mainly divided into ： Single transaction multi-user test and mixed transaction multi-user test , Because you have to run stability in the end , This time, only single transaction multi-user test is conducted .

Each typical transaction is executed through single transaction and multi-user iteration , Get performance metrics , such as TPS、ART、 System resource usage , Tune performance as needed .

tps When there is an inflection point , Continue to measure 2 Group data , If this 2 Group data tps Obvious decline , At this point, the maximum number of concurrent users is measured .

remarks ： The transaction database has current flow table and historical flow table , Therefore, delete the data of the current flow table before each scenario run .

2.3 Stability test

Multi transaction and multi-user concurrent hybrid mode , Carry out long-term stability test on the tested system , Obtain the performance index under continuous pressure . Investigate whether there will be downtime 、 Response time is getting longer 、 The success rate of the transaction decreased 、 The utilization rate of resources reaches 99% The situation of .

Select the maximum number of concurrent users when a single transaction is concurrent, and take the intermediate value .

3. Test summary

3.1 Bottlenecks and solutions

1、 The disk exchange rate has reached 99%;

2、 Memory usage 5% about ; Increase the number of system processes and concurrent users .

3、 Memory usage up to 99%; After checking the real-time log of the system , The reason is that some optional parameter is not configured .

4、 Single transaction sql execution time 2s; Add index .

5、 The execution time of single transaction is too long ; Every time sum Amount when , Too many transactions , Execution time is too long , Add a table for each transaction sum once , Shared the burden .

6、tps The value decreases significantly at some point in time , After query, the current flow table data is greater than 100 Ten thousand ; At present, there is no solution to tuning the database .